Hi,

I have done differential expression analysis with DESeq2. With goseq I can see there are pathways of interest differentially expressed- that makes sense to our biological observations. What I would like to do next is self-contained gene set testing with molecular pathway signature from MSigDB. Many answers here seem to suggest that, roast is best suited to follow edgeR or limma DGE analysis.

My question is: is there a reasonable method to do the similar self-contained test with DESeq DE genes or do I have to do the whole analysis again in edgeR ?

Thank you

With RNA-seq data, I'd suggest you use fry(), which is equivalent to roast in this context but with infinitely many rotations.

fry() and roast() work on the raw counts, not on a list of DE genes. They give results that are not directly related to any genewise DE analysis but they do use estimated NB parameter values. I suppose that fry() could be rewritten to use model parameters out of DESeq2 but it wouldn't be straightforward and I haven't done that. So yes you would need to use edgeR. Or alternatively, export some sort of log expression values out of DESeq2 and use limma.

You could consider using cameraPR, which is a competitive test but can accept a test statistic from any source including DESeq2. If you plan on testing large numbers of MSigDB pathways, I'd recommend camera anyway over roast for fry.